Working Paper: NBER ID: w22401
Authors: Pelin Akyol; James Key; Kala Krishna
Abstract: We model and estimate the decision to answer questions in multiple choice tests with negative marking. Our focus is on the trade-off between precision and fairness. Negative marking reduces guessing, thereby increasing accuracy considerably. However, it reduces the expected score of the more risk averse, discriminating against them. Using data from the Turkish University Entrance Exam, we find that students' attitudes towards risk differ according to their gender and ability. Women and those with high ability are significantly more risk averse: nevertheless, the impact on scores of such differences is small, making a case for negative marking.
Keywords: multiple choice exams; negative marking; risk aversion; gender differences; test performance
JEL Codes: I21; J24; D61; C11
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| eliminating negative marking (C20) | women's representation in top score distributions (C46) |
| negative marking (C52) | students' test-taking behavior (C92) |
| students' test-taking behavior (C92) | accuracy (C52) |
| risk aversion (D81) | guessing behavior (C92) |
| risk aversion (D81) | skipping questions (C34) |
| skipping questions (C34) | gender gap in exam performance (I24) |
| risk preferences (D81) | gender gap in exam performance (I24) |